Control of Ammonium Concentration in Escherichia coli Fermentations

B. G. Thompson, M. Kole and D. F. Gecson* Biotechnology Department, Alberta Research Council, 11315 87th Avenue, Edmonton, Alberta, T6G 2C2 Canada

Accepted for publication September 27, 1984

A control system has been devised for the maintenance of stable ammonium concentrations throughout a fed- batch fermentation. The control system is based on an ammonia gas-sensing electrode that monitors a pH-ad- justed effluent stream from the fermentor. To overcome the time lag between the fermentor and the electrode, feedback control included metered flows of ammonium to both the fermentor and the electrode vessel. The system was used to study the growth of Escherichia coli B (ATCC 11303) at controlled ammonium concentrations of 5 to 200mM. Apparent specific growth rates, biomass and protein production, and glucose yields were essentially constant from 5 to 170mM. Above 170mM ammonium growth was inhibited. As ammonium concentration de- creased from 170 to 5mM, ammonium yields increased from 1 to 24 g cell dry wt/g ammonium utilized. The results demonstrate that control of ammonium concen- trations at levels so low that ammonium would be ex- hausted in batch fermentations can significantly increase overall ammonium yields.

INTRODUCTION

There are two major routes for the assimilation of ammonium by Escherichia coli. At low ammonium concentrations, typically less than ImM, glutamate is aminated by glutamine synthetase' to form glutamine. High ammonium concentrations inhibit the synthesis of glutamine synthetase and stimulate the synthesis of glutamate by glutamate dehydrogenase.* Both glu- tamine and glutamate are key metabolites in amino acid biosynthesis, and their availability affects the flow of assimilated ammonium into protein.

In order to further delineate the effects of ammonium concentration on its assimilation by E. coli, we felt it would be useful to examine growth and ammonium assimilation under controlled ammonium concentra- tions. To accomplish this, we have modified the am- monium control system described by Hill and Thom- ~ n e l . ~ They developed an ammonium-monitoring system based on an ammonia gas electrode, but their attempts to use it for control resulted in oscillations in ammonium concentration. The control system described here maintains more stable ammonium concentrations, al-

* To whom all correspondence should be addressed.

Biotechnology and Bioengineering, Vol. XXVII, Pp. 818-824 (1985) 0 1985 John Wiley & Sons, Inc.

lowing the study of the effect of controlled ammonium concentrations on ammonium assimilation in E. coli. The studies presented here are concerned with the effects of relatively high ammonium concentrations (5-200mM), since over this concentration range few changes in enzyme expression take place, and greatest stability could be achieved in the ammonium control system.

MATERIALS AND METHODS

Microorganism, Media, and Fermentations

The organism used in this study was E. coli B (Luria) ATCC 11303. It was maintained at room temperature (25 ? 2C) on a modified medium based on that of Davis and M i n g i ~ l i . ~ The basal medium used for all fermentations was the modified Davis and Mingioli medium with reduced buffering capacity and the fol- lowing composition: MgS04 * 7H20, 0.1 g/L; K2HP04, 1.5 g/L; NaH2P04, 0.25 g/L; glucose, 8.8 g/L; and variable amounts of (NH4)*S04, at pH 7.0. For inocula, maintenance cultures, and all pH-con- trolled, ammonium-uncontrolled fermentations, the (NH4)$04 concentration was 2.0 g/L. Fermentations were performed in a New Brunswick Multigen fer- mentor (New Brunswick Scientific, New Brunswick, NJ) in a 2-L vessel containing (initially) 1.75 L medium (Fig. 1) . Temperature was controlled at 37"C, mixing was with two 6-bladed flat-blade impellers operating at 250 rpm and an aeration rate of 0.40 L air (L liquid min)-'. The oxygen transfer rate under these conditions was found to be 22mM/L h by purging the tank with oxygen-free N2 and recording the reoxygenation of complete, but cell-free, medium with an IL 530 oxygen electrode system (Instrumentation Laboratories, An- dover, MA). The pH of the fermentation broth was maintained by a Chemcadet pH/millivolt controller (Cole-Parmer, Chicago, IL) which supplied various titrants (see below) by on-off activation of Masterflex peristaltic pumps (Cole-Parmer, Chicago, IL) (Fig. 1, controller A).

The effect of ammonium, as ammonium chloride,

CCC 0006-3592/85/060818-07$04.00

DUAL CHART

NH4+ @ CONTROLLER RECORDER 1

ACID

BASE

5 2 1

m CONTROLLER

IJ @

Figure 1. Ammonium-controlled fermentation system. Controller A maintains pH in fermentation vessel. Controller B responds to ammonium concentration sensed by ammonia electrode (b) in the measurement vessel by pumping ammonium sulfate into both fermentation vessel (flow 3) and measurement vessel (flow 5). Controller C maintains pH in measurement vessel between pH 11.0 and 11.5 to satisfy requirements of ammonia electrode.

tJ-l-z-j------

on shake-flask cultures of E. coli was determined as follows. 100-mL shake-flask cultures in 250-mL flasks containing 2 g/L (NH4)*S04 were grown into early log phase. At this point 25 mL containing the amount of NH4Cl required to achieve the desired final am- monium concentration were added to the shake flask. No attempt was made to distinguish the relative effects of ammonium and chloride in this experiment. The cultures were allowed to grow for another 6 h, and the relative increase in biomass over that period was determined.

Inocula for all fermentations were 100-mL shake- flask cultures grown to stationary phase (overnight) on a New Brunswick G25 gyrorotary shaker at 180 rpm and 37C. Addition of the entire inoculum resulted in initial biomass concentrations of 0.006-0.01 g/L. Regardless of the titrant combination used, the pH of the fermentation broth was maintained between pH 6.70 and pH 6.90. The various titrants used during pH-controlled, ammonium-uncontrolled fermentations were (a) 0.1N phosphoric acid and 35% (w/v) Na2HP0,, (b) 0.1N HCI and 40% (w/v) NH40H, or (c) 0.1N HCl and 2N NaOH. The titrant combination used during pH-controlled, ammonium-controlled fer- mentations was 0.1N HCl and 2N NaOH. Samples were taken every hour throughout each fermentation. Biomass was determined spectrophotometrically (OD,,) and converted to grams cell dry wt by com- parison to a standard conversion curve constructed for this strain of E. coli. Total glucose in the cell-free medium was determined by the method of Dubois et

ELECTRODE

aL5 Cells were removed from the culture broths either by filtration (0.22 p m pore diameter Millex filter units, Millipore Corp., Bedford, MA) or by centrifugation (Brinkman Eppendorf, model 5412, at 15,OOOg for 3 min in 1.5-mL tubes). The ammonium concentration of samples was determined with an Orion model 95- 10 ammonia gas electrode (Orion, Cambridge, MA) using the manufacturers recommended sample prep- aration and calibration techniques for various am- monium concentrations in complete medium. Protein was determined by the Bio-Rad method6 (Bio-Rad, Control No. 26402). Ion concentrations Na, SO:-, Mg2+, and PO:- were determined by the Chemical Analysis Group, Alberta Research Council, with an Argon plasma ICP spectrometer [ARL 34,0001.

Amino acids excreted into the growth medium during the growth of E. coli were detected as follows. The cell free supernatant was dried under air and resus- pended in one-half the original volume of distilled water. Samples were spotted onto silica gel H plates and the chromatograms were developed in I-butanol-acetic acid-water (4 : 1 : 1 v/v). After the plates had dried thoroughly, amino-containing compounds were detected with 0.25% (v/v) ninhydrin in acetone.8 Proline, glu- tamine, and glutamate were used as standards.

Yields for glucose and ammonium are reported as biomass yields for each substrate ( Yx,s) and are given in units of g cell dry wt/g substrate consumed in the course of the entire fermentation (neither lag nor sta- tionary phase were of sufficient duration to warrant their exclusion).

THOMPSON, KOLE, AND GERSON: CONTROL OF AMMONIUM CONCENTRATION 819

Control of Ammonium Concentration

The concentration of ammonium in the fermentor was controlled by the system diagrammed in Figure 1. A Chemcadet pH/millivolt controller was used in conjunction with the ammonia gas electrode to de- termine the ammonium concentration in the effluent stream from the fermentor (Fig. 1, controller B). Control was by on-off action. The ammonia gas electrode must be operated between pH 11.0 and pH 11.5 to ensure that all of the ammonium ion present in the sample is converted to ammonia gas. This requirement led to the development of the following control system.

Whole fermentation broth was continuously removed at a rate of 30 mL/h from the fermentor by a FMI model G-150 metering pump operated by a stepping motor (FMI, Oyster Bay, NY) (Fig. 1, flow 4). This passed through a drop-gap to prevent back-flow or contamination of the fermentor into a measurement vessel that contained the ammonia gas electrode (Fig. 1, electrode b). The measurement vessel was controlled between pH 1 1 .O and pH 11.5 by a Radiometer model TTTlA pH controller (Radiometer, Copenhagen, Den- mark) (Fig. 1, controller C), and ammonia gas escape was prevented by covering the measurement vessel with Parafilm. The level of the measurement vessel was maintained constant (at a volume of 10 mL) by continual withdrawal with a Masterflex peristaltic pump (Cole-Parmer Instrument Co.) (Fig. 1 , flow 7). Control of the ammonium concentration in the fermentor was effected by the simultaneous operation of two pumps (Fig. 1, flows 3 and 5 ) when the electrode output indicated that the ammonium concentration in the measurement vessel had decreased below the set point. A Masterflex model No. 7020-40 peristaltic pump de- livered (NH4)*S04 to the fermentation vessel and a Sage model 341A syringe pump delivered (NH&S04 to the measurement vessel. The reason for this ar- rangement was to overcome the time delay of ap- proximately 4.7 min in the delivery of whole fermen- tation broth to the measurement vessel (Fig. 1, flow

4). This time delay resulted in significant oscillations in the ammonium concentration in the measurement vessel and the fermentor (Fig. 2B). Inclusion of the corrective flow to the measurement vessel reduced the oscillations in ammonium concentration to an am- plitude similar to those observed in pH control (Figures 2A and C).

In the operation of the ammonium control system, the concentration of (NH4)2S04 supplied to the fer- mentation vessel was 5 times the set-point concen- tration, and the concentration of (NH,),SO, supplied to the measurement vessel was 10 times the set-point concentration. Successful operation of such a control system requires judicious selection of the various pumping rates (Fig. 1, flows 3,5). Inappropriate pump- rates could result in control of the ammonium con- centration in the measurement vessel while the am- monium concentration in the fermentor increased, de- creased, or fluctuated. Pumping rates were determined empirically. Throughout each fermentation the con- centrations observed in the measuring vessel were compared to the concentrations of ammonium present in samples taken directly from the fermentation vessel.

RESULTS

pH Control Strategies

The results of 2-L fermentations conducted with various pH control strategies are summarized in Figure 3 and Table I. The use of either NaOH/HCl or Na2HPO4/H3PO4 as the titrant combination gave the highest cell densities (2.8 g cell dry wt/L) and essentially identical apparent maximum specific growth rates ( p = 1.0 h-I). Control of pH with the NH,OH/HCl titrant combination did not affect the apparent maximum specific growth rate but slowed growth late in the fermentation and reduced the ultimate cell density to 1.5 g cell dry wt/L. As expected, the biomass pro- duction of cultures grown without pH control was

7.05 - pH 6.75 -

6.45 - P A

C 18 -

NH4+(mM) 14 - I I 1 I I I I I 3 0 1 2 3 4 5 6 7 8

Time (H)

Figure 2. Simultaneous traces of pH in fermentation vessel (trace A) and ammonium concentrations (mM) in measurement vessel over 8-h period (traces B and C). Trace B gives ammonium concentration in measuement vessel when there was simple feedback control. Trace C gives ammonium concentration in measurement vessel when there was feedback control with compensatory control to measurement vessel. Set points of B and C were the same.

820 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 27, JUNE 1985

4.0 I I I I I I

0.001 2 4 6 8 1 0

Time [hl

Figure 3. Growth curves of E. coli in 2-L batch fermentations with various control strategies. (0) Titration with Na,HPO,/H,PO, ; (0) titration with NaOH/HCI; (m) titration with NH,OH/HCI; (A) pH uncontrolled, ammonium uncontrolled; (0) pH uncontrolled, am- monium controlled.

greatly reduced by acidification. This effect was only slightly ameliorated by the provision of ammonium control. When pH was controlled using NH40H as titrant, the concentration of ammonium reached 180mM,

by ammonium excess. Glucose yields (g cell dry wt/g glucose) were highest when pH was controlled by NaOH/HCl or Na2HP04/H,P04 and were significantly decreased in the absence of pH control. The excess ammonium provided by pH control with NH40H slightly depressed glucose yields, but control of am- monium at ca. 0.1M slightly increased the glucose yield in the absence of pH control.

Ammonium Inhibition

Since the results of the pH control experiments sug- gested inhibition by excessive ammonium concentra- tions, the effect of ammonium concentration was further investigated. The results of shake-flask studies of the effect of initial ammonium concentration on ultimate biomass production are given in Figure 4. Under these conditions (e.g., the absence of pH control), a sharp optimum ammonium chloride concentration (200mM) was found for biomass production. Ammonium chloride concentrations above 480mM completely inhibited growth. These findings supported the suggestion of ammonium toxicity resulting from the effects of pH control with NH40H but did not exclude possible effects of chloride. The resuIts of these experiments suggested that there may be benefits from the control of am- monium concentrations throughout the fermentation.

Ammonium Control

and it appears that this ammonium concentration is inhibitory.

Yields under the three pH control strategies were also determined. Ammonium yields (g cell dry wt/g ammonium) were sharply depressed (Table I) by the addition of excess ammonium during pH control. Am- monium yields were less affected by acidification than

An example of the time course for a fermentation in which the concentration of ammonium ion was maintained at a constant level is given in Figure 5 . The set point for the ammonium control in this case was 20mM ammonium. Ammonium concentrations in samples taken from the fermentor are given in Figure 5 and are relatively constant over the 8-h period. Fluc-

Table I. Effects of pH and ammonium control on E. coli fermentations.

pH control No Yes Yes Yes No Ammonium control No No No No Yes Titrant s

Base - NaOH Na,HPO, NH,OH - Acid - HCI H,PO, HCI -

Biomass produced,

Ammonium utilized, g/L 0.098 0.44 0.47 1.02 1.02

Glucose utilized, g/L 2.05 8.41 8.45 6.35 2.45 Y (ammonium) 3.98 6.36 5.% 1.51 4.24 Y (glucose) 0.190 0.333 0.331 0.244 0.250 P- 1.09 1.02 1.04 0.976 1 .00 Final values

Na g/L - 2.33 6.30 0.49 0.05 Phosphate g/L - .69 11.49 0.91 0.9 Sulfate g/L - 1.46 1.13 1.52 3.36 Ammonium, mM 24.9 5.9 4.7 180.0 10.8 PH 4.30 6.80 6.80 6.80 3.90

g cell dry wt/L 0.39 2.80 2.80 1.55 0.615

Total Ammonium (M)

Figure 4. Shake-flask study of effect of ammonium concentration on relative biomass increase.

THOMPSON, KOLE, AND GERSON: CONTROL OF AMMONIUM CONCENTRATION 82 1

10

1 .o

+ - 0.1 -I . En - ii E I

.01

.001

II - 3

!OOO f - X 2

500 9 - P - b

8

6

0 2 4 6 8 Time (h)

- -I 6

5 a li p 4 f 3 - 1 2 -

1 -

0 -

Figure 5. Time course of ammonium-controlled fermentation. Ammonium controlled at approximately 20mM. (+) Biomass (g cell dry wt/L); (0) glucose concentration (g/L); (V) ammonium concentration in samples (M); (m) Na' concentration (mg/L); (0) magnesium concentration (mg/L); (A) sulfate concentration (mg/L).

2 0 - ,

7 - 3 3

-.C 8

. . - a 3

-% - 1.0 -

0 -

tuations of ammonium concentrations in the samples of about +5% of the set point were typical. Under these conditions the apparent maximum specific growth rate was 1 .O hr-', and the ultimate biomass production was 2.4 g cell dry wt/L. The buildup of SO:- throughout the fermentation reflects the cumulative addition of ammonium sulfate required to maintain a relatively constant ammonium concentration, while the buildup of Na' reflects the cumulative NaOH additions required to maintain constant pH. Phosphate concentrations in the medium declined only moderately, from 1 .O to 0.8 g/L. Glucose consumption followed growth, and glu- cose was essentially exhausted after 8 h. Magnesium declined to approximately 8.5 mg/L by the end of the fermentation.

The results of a number of fermentations in which ammonium was controlled at various concentrations from 5 to 200mM are given in Figures 6 and 7. From 5 to 170mM ammonium ultimate biomass production declined slightly from 2.7 to 2.4 g cell dry wt/L (Fig. 6). Above 170mM ammonium biomass production dropped precipitously, as was observed in the shake- flask experiments (above). Over the range from 5 to 170mM ammonium protein production, glucose utili- zation, and acid production, as reflected in NaOH consumption, were relatively constant. Above 170mM ammonium all of these also dropped precipitously. Ammonium utilization, however, increased steadily over the ammonium concentration range that permitted growth. The exact fate of this nitrogen has not yet been determined, but thin-layer chromatography re-

vealed the production of several amino acids which were excreted into the medium (see below).

The apparent maximum specific growth rate, F , was quite constant at 1.0 h-' up to the inhibitory ammonium concentration (Fig. 7). Protein as a percentage of bio- mass increased slightly (from 40 to 45%) with decreasing ammonium concentrations over this range. Glucose yields were also constant at approximately 0.3 g cell dry wt/g glucose consumed until the inhibitory con- centration of ammonium was reached.

2.0 t-I 1.6 1.4 1.0 f - loo $

0.6 - Bo f P

0.2 f - 20 ' 5 x 10'

Ammonium ion IMI

Figure 6. Effects of controlled ammonium levels on E. coli fer- mentations. (8) Glucose utilization (g/L); (V) cell protein (g/L); (+) biomass (g cell dry wt/L); (0) cumulative ammonium sulfate utilized for ammonium control (g/L); (A) cumulative NaOH utilized for pH control (mM/L).

822 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 27, JUNE 1985

0.4 - . x c I

0.3 - E x

0.2 -

0.1 -

0 - I I I 1 5 x 1 0 - ~ 1 0 . ~ z X 10.'5x 1 0 . ~ 10.' z X 10.'

Ammonium ion [MI

Figure 7. Effects of controlled ammonium levels on E. coli fer- mentations. (V) Ammonium yields (g cell dry wt/g ammonium); (+) glucose yields (g cell dry wt/g glucose); (0) apparent maximum specific growth rates (h-'); (W) protein/biomass ratio.

Ammonium yields, however, increased from 1 to 24 g cell dry wt/g ammonium as ammonium concen- tration decreased. This result is in agreement with the results obtained in the pH-controlled fermentations presented in Table I. The sharply decreasing ammonium yields with increasing ammonium concentrations are in accordance with the finding that soluble nitrogenous compounds (e.g., amino acids) were being produced at higher ammonium concentrations.

Using the solvent system described above, proline, glutamine, and glutamate have Rfvalues of 0.325,0.293, and 0.347, respectively. Four major ninhydrin-positive bands with Rf values of 0.007, 0.160, 0.358, and 0.507 were detected in the sample. The corresponding amino acids have not yet been determined.

DISCUSSION

For a variety of reasons, such as sensitivity to steam sterilization, many sophisticated sensors are difficult to insert directly into fermentation vessels. In the case at hand the requirement for high pH precludes use of the ammonia gas electrode for ammonium determi- nations in the fermentor. Such sensors can, however, be used in sample streams continuously removed from the fermentor for conditioning or separated by a sterility barrier (e.g., cross-flow filtration). In all such systems there will be a time delay in sensing the concentration of materials in the fermentor. This may not cause significant difficulties if the obective is simple data acquisition, but these delays could cause considerable difficulties if the information from the sensor is used for feedback control.

The ammonium control system used in this study has the peculiar characteristic that both the fermentation vessel and the sensor are the objects of feedback con-

trol. This control strategy is one way in which the effects of time delay in sensing can be overcome in a feedback control loop. The consequences of simul- taneously feeding the vessel, which is the object of control, and the obligatorily separated sensor are com- plex but can include damping of the oscillation caused by the time delay.

It is instructive to follow the course of events in a single cycle of the on-off control of ammonium con- centration, first, with simple feedback control and, second, with the feed-forward control system presented here. The sequence of events with simple feedback control and a set point of 5mM is as follows: (1) At some time, t = 0, the set point is reached in the fermentor; (2) 5 min later fluid at the set point reaches the measurement vessel; (3) if the washout time for the measurement vessel plus the electrode response is 10 min, the sensor triggers the controller at t = 15 min, by which time the concentration in the fermentor is 4.91mM; (4) the pump turns on, delivering 0.14 mmol/min, and bringing the fermentor to the set point in 0.6 min; (5 ) 5 min later fluid at the set point reaches the measurement vessel, and 10 min later (washout plus electrode response) the sensor detects the set point concentration; (6) because of the deadband of the controller, the pump is not turned off for another 2.5 min, by which time the concentration in the fer- mentor is 7.4mM; and (7) the resulting overshoot is approximately 48% of the set point. With feed-forward control such as we have described, the first time delay is unaltered, but at step 4 an additional pump delivers ammonium to the measurement vessel and takes 0.6 min (at the pump rate used) to turn the controller off, by which time the concentration within the fermentor is within 1-2% of the set point.

With experimental values for the ammonium con- sumption rate of the culture and for the time delays of the various components of the control system, it is possible to estimate the relative error of the simple feedback system and the pump rates of the feed-forward control system. With constant pumping rates one would expect the errors of the control system to be a function of cell concentration, but these errors will always be greater for the simple feedback control case than for the feed-forward control case. In both cases the fre- quency of operation of the controller will increase with increasing cell concentration.

Over the noninhibitory ammonium concentration range studied here, p for E. coli was relatively constant at 1.0 h-'. In the early stages of these fermentations both glucose and ammonium are in excess. Although glucose concentration was steadily declining (e.g., see Fig. 5), p remained constant until a biomass concen- tration of 2.5-2.8 g cell dry wt/L was reached. Growth yields for ammonium ( Yx/ammon,um) increased approxi- mately 25-fold as the level at which the ammonium concentration was controlled decreased from 170 to 5mM. The maximum Yxlammon,um in the presence of

THOMPSON, KOLE, AND GERSON: CONTROL OF AMMONIUM CONCENTRATION 823

ample glucose over this range was 24 g cell dry wt/g ammonium consumed, at an ammonium concentration of 5mM. At this concentration more ammonium was consumed in the controlled fermentation than woufd be present in a batch fermentation of the same initial ammonium concentration. Control at 20mM resulted in a Yxlammonrum of 10 g cell dry wt/g ammonium con- sumed, whereas at a steady-state ammonium concen- tration of 20mM, in glucose limitation, Senior obtained a Yxlammon,um of 3.3 g cell dry wt/g a rnmoni~m.~ Under ammonium-limiting, glucose-excess conditions, Senio? obtained a Yxlammonium of approximately 5.5 g cell dry wt/g ammonium consumed, which is similar to the results obtained in pH-controlled, ammonium-uncon- trolled fermentations (Table I). It thus appears that appropriately controlled concentrations of ammonium

in the presence of sufficient glucose can improve am- monium yields in E. coli fermentations.

References

1 . B . Magasanik, Ann. Rev. Genet., 16, 135 (1982). 2. J . A. Pateman, Biochem. J . , 115, 769 (1969). 3. F. F. Hill and J. Thommel, Proc. Biochem., 17(5), 16 (1982). 4. B . D. Davis and E. S. Mingioli, J . Bacteriol. 60, 17 (1950). 5. M. Dubois, K . A. Gilles, J. K. Hamilton, P. A. Gebers, and F.

6. M. Bradford, Ann. Biochem., 72, 248 (1976). 7. J . G. Kirchner, in Thin Layer Chromatography, Techniques of

Organic Chemistry, Vol. 12, E. S . Perry and A. Weissberger, Eds. (Wiley, New York, 1967).

8. M. Kates, in Techniques in Lipidology, T. W. Work and E. Work, Eds. (North-Holland Publishing Co., Oxford, U.K. , 1972).

9. P. J . Senior, J . Bacteriol, 123, 407 (1975).

Smith. Anal. Chem., 28, 350 (1956).

824 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 27, JUNE 1985